At 5:20 AM -0800 11/22/08, Jari Arkko wrote: > >*) Even if it should be understood that vendors make their own decisions >independently of the IETF, some of the similar application implications >may exist due to other reasons such as firewalls, etc. The warts due to firewalls are very different from those due to topology shifting from an application-layer perspective. I think the ICE and ICE-TCP documents ought to give everyone serious pause about creating NATs in v6 space; even with all the work ICE requires (a rendezvous protocol, relay servers, candidate gathering etc.) they do not work anywhere close to a 100% of the time. As originally specified, I seem to recall ice-tcp succeeded something like 40% in the real world. Tony's point is architectural, and I agree with it. The folks who are arguing that the practical trumps the architectural here are also very likely to be wrong, as the practical consequences here are painful indeed. regards, Ted _______________________________________________ Ietf@xxxxxxxx https://www.ietf.org/mailman/listinfo/ietf